Aiming at the problem of low resource utilization and high energy consumption in cloud data center, a resource balancing scheduling strategy based on the resource demand difference was proposed. Based on a package-cluster framework model, the packages with large differences in resource requirements were clustered an improved k-means algorithm using the distance metrics related to resource requirements. The resources were used as the distance between packages and clusters. In the process of resource allocation, the package was mapped into clusters in a centralized manner, thereby the number of clusters used could be reduced. The experimental results show that under the concept of package-cluster framework, the improved k-means clustering algorithm based on the difference of resource requirements can optimize the packet clustering step. The resource scheduling algorithm presented can improve the utilization of various resources and reduce the energy consumption in the cloud data center. The algorithm is of effectiveness and scalability.